Empty Container Verification Using Deep Learning
Wei Shi, Cun Cheng, Zhiyuan Luo, Yanjie Yao, Yu-Ying Hong · 2017
The inspection of X-ray cargo images is a challenging visual search task for security department. With the increase of import and export trade, it is necessary to develop an intelligent algorithm to improve the accuracy and efficiency of inspection by assisting security inspectors to check the cargo. This paper is proposed a method to solve the problem of empty container verification. According to characteristics of containers, we solve this task in two steps. Firstly, we use rule-based algorithm to locate the container position in the X-ray image. And then deep learning method is applied to identify containers that have been segmented. Finally, accuracy of the proposed method achieves 99.86% in the test set, 1.42% higher the Faster-RCNN. And test speed is 3 times faster than Faster-RCNN.